Portail HAL edf
Not a member yet
11029 research outputs found
Sort by
A Bandit Approach with Evolutionary Operators for Model Selection: Application to Neural Architecture Optimization for Image Classification
No full textThis work formulates model selection as an infinite-armed bandit problem, namely, a problem in which a decision maker iteratively selects one of an infinite number of fixed choices (i.e., arms) when the properties of each choice are only partially known at the time of allocation and may become better understood over time, via the attainment of rewards.Here, the arms are machine learning models to train and selecting an arm corresponds to a partial training of the model (resource allocation).The reward is the accuracy of the selected model after its partial training.We aim to identify the best model at the end of a finite number of resource allocations and thus consider the best arm identification setup. We propose the algorithm Mutant-UCB that incorporates operators from evolutionary algorithms into the UCB-E (Upper Confidence Bound Exploration) bandit algorithm introduced by Audiber et al.Tests carried out on three open source image classification data sets attest to the relevance of this novel combining approach, which outperforms the state-of-the-art for a fixed budget
A Bayesian Inference Approach to Extract Circuit Model Parameters and Analyze Photovoltaic Degradation from Power Production Data
No full textInternational audienc
Multifractal analysis of aerosol particle concentration during rain and dry conditions in nm and µm range
No full textInternational audienceBelow cloud scavenging by rain is known to be a very efficient sinking mechanism for aerosols in atmosphere. Since this scavenging depends on interaction between aerosol particles as well as the scavening raindrops, and notably their respective size ranges, it is interesting to examine both fields together across various size ranges and across temporal scales. Towards this, a 4 month long data was used from Cherbourg-Octeville, France from 01/11/2010 to 12/03/2011 from the experimental station managed by Institut de Radioprotection et de Sûreté Nucléaire (IRSN). Here, simultaneous and continuous measurement of size resolved particle concentration (14.6 to 478.3 nm and 0.523 to 19.81 µm) range has been done using Scanning Mobility Particle Sizer (SMPS) and Aerodynamic Particle Sizer (APS), and rain measurement using a disdrometer. Variation of total aerosol concentration in nm and µm range, as well as individual number concentration in small size bins were analyzed according to rain and dry events, using the framework of Universal Multifractals (UM). UM is widely used, as a physically based scale invariant framework, for characterizing and simulating extreme variability and intermittency in geophysical fields. From initial analysis, the total concentration showed scaling properties (1 min to 1 hr), in both rain and dry events, regardless the scavenging efficiency of event. This was further explored in individual concentration ranges and they showed similar scaling properties in different rain types. However, while considering the different stages of rain, say start and end, the values of UM parameters showed some variation. To understand the behavior more clearly, few sizes were selected from nm and µm range, and efforts were made to extract the field which is devoid of scavenging by rain. Understanding the correct transformation required to extract accurate UM values and comparing the scavenging and non scavenging fields will improve understanding of particle concentration variation, and eventually understanding of scavenging coefficient
Multidomain FEM-BEM coupling for acoustic scattering
No full textInternational audienceWe model time-harmonic acoustic scattering by an object composed of piece-wise homogeneous parts and an arbitrarily heterogeneous part. We propose and analyze new formulations that couple, adopting a Costabel-type approach, boundary integral equations for the homogeneous subdomains with volume variational formulations for the heterogeneous subdomain. This is an extension of the Costabel FEM-BEM coupling to a multi-domain configuration, with cross-points allowed, i.e. points where three or more subdomains are adjacent. While generally just the exterior unbounded subdomain is treated with the BEM, here we wish to exploit the advantages of BEM whenever it is applicable, that is, for all the homogeneous parts of the scattering object. Our formulation is based on the multi-trace formalism, which initially was introduced for acoustic scattering by piece-wise homogeneous objects. Instead, here we allow the wavenumber to vary arbitrarily in a part of the domain. We prove that the bilinear form associated with the proposed formulation satisfies a Gårding coercivity inequality, which ensures stability of the variational problem if it is uniquely solvable. We identify conditions for injectivity and construct modified versions immune to spurious resonances
Physical insight into interactions of interstitial loops and dislocation lines in austenitic high entropy alloys: atomic-scale modelling
No full textInternational audienceNuclear reactor materials undergo significant changes in their microstructure and mechanical properties due to radiation exposure. Understanding the role of radiation-induced defects in these materials is a complex challenge, especially for new candidate material. This study utilizes molecular dynamics (MD) simulations with a novel interatomic potential to investigate the interactions between mobile edge dislocation lines and interstitial Frank loops within the quaternary model austenitic high entropy alloy (HEA) Cr15Fe46Mn17Ni22. This candidate alloy is referenced as "Y3-HEA". The behavior of this alloy is compared to that of a ternary alloy (Cr20Fe70Ni10) considered as a model of classical austenitic stainless steel (ASS).The investigation begins with molecular statics analysis, focusing on the unfaulting process of interstitial Frank loops. A simplified thermodynamic model is introduced to predict the critical size at which a Frank loop can undergo unfaulting. Complementing this, MD simulations explore several variables, including loop orientation, random seed configuration, temperature, loop size, and intersection interactions. Through this examination, interaction mechanisms between mobile edge dislocations and interstitial Frank loops in Y3-HEA are classified and comparisons are drawn with those observed in ASS.Furthermore, an average obstacle strength parameter is calculated for both ASS and Y3-HEA and its value is integrated into a radiation hardening constitutive model. This approach allows for the prediction of the contribution of dislocation loops to the hardening behavior of the studied alloys
Growth dynamic of biofilm-associated Naegleria fowleri in freshwater on various materials
No full textInternational audienceIn industrial water systems, the occurrence of biofilm-associated pathogenic free-living amoebae (FLA) such as Naegleria fowleri is a potential hygienic problem, and factors associated with its occurrence remain poorly understood. This study aimed to evaluate the impact of four cooling circuit materials on the growth of N. fowleri in a freshwater biofilm formed at 42°C and under a hydrodynamic shear rate of 17 s −1 (laminar flow): polyvinyl chloride, stainless steel, brass, and titanium. Colonization of the freshwater biofilms by N. fowleri was found to be effective on polyvinyl chloride, stainless steel, and titanium. For these three materials, the ratio of (bacterial prey)/(amoeba) was found to control the growth of N. fowleri . All materials taken together, a maximum specific growth rate of 0.18 ± 0.07 h −1 was associated with a generation time of ~4 h. In contrast, no significant colonization of N. fowleri was found on brass. Therefore, the contribution of copper is strongly suspected
Simultaneous off-the-grid learning of mixtures issued from a continuous dictionary
No full textIn this paper we observe a set, possibly a continuum, of signals corrupted by noise. Each signal is a finite mixture of an unknown number of features belonging to a continuous dictionary. The continuous dictionary is parametrized by a real non-linear parameter. We shall assume that the signals share an underlying structure by assuming that each signal has its active features included in a finite and sparse set. We formulate regularized optimization problem to estimate simultaneously the linear coefficients in the mixtures and the non-linear parameters of the features. The optimization problem is composed of a data fidelity term and a -penalty. We call its solution the Group-Nonlinear-Lasso and provide high probability bounds on the prediction error using certificate functions. Following recent works on the geometry of off-the-grid methods, we show that such functions can be constructed provided the parameters of the active features are pairwise separated by a constant with respect to a Riemannian metric.When the number of signals is finite and the noise is assumed Gaussian, we give refinements of our results for and using tail bounds on suprema of Gaussian and random processes. When , our prediction error reaches the rates obtained by the Group-Lasso estimator in the multi-task linear regression model. Furthermore, for these prediction rates are faster than for when all signals share most of the non-linear parameters
Energy depletion by cell proliferation sensitizes the kidney epithelial cells to injury
No full textInternational audienceAcute kidney injury activates both proliferative and anti-proliferative pathways, the consequences of which are not fully elucidated. If an initial proliferation of the renal epithelium is necessary for the successful repair, the persistence of proliferation markers is associated with the occurrence of chronic kidney disease. We hypothesized that proliferation in stress conditions impacts cell viability and renal outcomes. We found that proliferation is associated with cell death after various stresses in kidney cells. In vitro, the ATP/ADP ratio oscillates reproducibly throughout the cell cycle, and cell proliferation is associated with a decreased intracellular ATP/ADP ratio. In vivo, transcriptomic data from transplanted kidneys revealed that proliferation was strongly associated with a decrease in the expression of the mitochondria-encoded genes of the oxidative phosphorylation pathway, but not of the nuclear-encoded ones. These observations suggest that mitochondrial function is a limiting factor for energy production in proliferative kidney cells after injury. The association of increased proliferation and decreased mitochondrial function was indeed associated with poor renal outcomes. In summary, proliferation is an energy demanding process impairing the cellular ability to cope with an injury, highlighting proliferative repair and metabolic recovery as indispensable and interdependent features for successful kidney repair
Projection-based model order reduction for prestressed concrete with an application to the standard section of a nuclear containment building
No full textWe propose a projection-based model order reduction procedure for the ageing of large prestressed concrete structures. Our work is motivated by applications in the nuclear industry, particularly in the simulation of containment buildings. Such numerical simulations involve a multi-modeling approach: a three-dimensional nonlinear thermo-hydro-visco-elastic rheological model is used for concrete; and prestressing cables are described by a one-dimensional linear thermo-elastic behavior. A kinematic linkage is performed in order to connect the concrete nodes and the steel nodes: coincident points in each material are assumed to have the same displacement. We develop an adaptive algorithm based on a Proper Orthogonal Decomposition (POD) in time and greedy in parameter to build a reduced order model (ROM). The nonlinearity of the operator entails that the computational cost of the ROM assembly scales with the size of the high-fidelity model. We develop an hyper-reduction strategy based on empirical quadrature to bypass this computational bottleneck: our approach relies on the construction of a reduced mesh to speed up online assembly costs of the ROM. We provide numerical results for a standard section of a double-walled containment building using a qualified and broadly-used industrial grade finite element solver for structural mechanics (codeaster)